China's Unitree says "GPT moment" for robots remains years away

Robotics may still be two to five years away from a “GPT moment” comparable with ChatGPT’s breakthrough in generative artificial intelligence, a senior executive at one of the leading Chinese humanoid makers said on July 22.

Irving Chen, director for the Asia Pacific and other areas at Unitree Robotics, said large language models (LLMs) had demonstrated their abilities in software but remained poorly connected to robots and machines that need to operate in the physical world.

“The LLM is powerful in the virtual world,” Chen told Nikkei‘s Global Digital Summit 2026 in Tokyo, but it had yet to show the same capability when controlling hardware, which needs to process more variables to operate in different environments.

Chen’s comments highlight the challenges facing the booming humanoid robotics industry. Robots are expected to bring autonomous physical AI into factories, but most remain confined to entertainment and demonstrations based on preset movements.

The Hangzhou-based startup has become one of the most high-profile companies in the humanoid robotics space, with its viral synchronized dances and martial arts displays in China. Chen said the performances show that the hardware is becoming reliable and scalable, but has not yet demonstrated intelligence suited to productive work.

For Unitree, commercial demand remains concentrated in research and testing. In the first nine months of last year, 74% of its humanoid robot revenue came from research and education, with just 9% from industrial customers.

Chen said wider adoption would require robots to perform longer and more complex tasks reliably in changing environments without pre-programming. “Right now, most robotics projects are in the pilot stage,” he said.

Unitree expects proof-of-concept trials to dominate through next year, followed by commercial deployment by 2030. From then on, robots could begin assisting humans with a broader range of general labor tasks, Chen said.

He identified more capable robot “brains” and large-scale manufacturing for the industry’s next breakthrough. Software models need to interpret their surroundings and turn instructions into physical actions, with faster computing response times necessary. Manufacturers have to produce durable machines at lower cost and in larger volumes.

Unitree, founded in 2016, is preparing for an IPO in Shanghai after receiving regulatory approval earlier this month. It plans to raise RMB 4.2 billion (USD 619.7 million), with 48% allocated to AI and embodied intelligence research.

A further RMB 1.1 billion (USD 162.3 million), or over 26%, would go toward newer hardware development, while RMB 624 million (USD 92.1 million), close to 15%, would be used for new factories with capacity to produce around 20,000 robots a year. The company shipped 6,500 units last year.

He added that the company had increased investment in embodied intelligence, which refers to the integration of AI into physical systems, and plans to launch more robotics models designed to work across homes, offices and industrial sites without pre-setting.

“Our expectation is one model that can fit most scenarios,” Chen said.

This article first appeared on Nikkei Asia. It has been republished here as part of 36Kr’s ongoing partnership with Nikkei.

Note: RMB figures are converted to USD at rates of RMB 6.78 = USD 1 based on estimates as of July 27, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.